Employees are embracing BYOAI – Bring Your Own AI to work, with or without employer approval. Employers still suffering from a lack of skills. Dell Plans to colour code employees based on office attendance. And the story continues about that mysterious GPT-2 chat bot.
All this and more on the “rumours and research reports” edition of Hashtag Trending. I’m your host, Jim Love. Let’s get into it.
A new joint research report from Microsoft and LinkedIn shows employees are enthusiastically embracing AI to help them in their jobs – and they aren’t waiting around for their companies to provide the tools. An overwhelming 75% of knowledge workers globally report using generative AI at work. But this usage is often bringing their own AI tools, with 78% saying they are pursuing a “bring your own AI” or BYOAI approach.
Meanwhile, there appears to be some hesitancy and lack of strategic vision from business leadership. While 79% of leaders believe AI adoption is critical for staying competitive, a full 60% worry their organization lacks a clear implementation plan and vision for AI.
The reports notes the rise of what it calls “AI power users”. These are employees using AI several times per week or more, with 92% saying it boosts their creativity and 93% feeling it allows them to focus on higher priority work. Power users report saving over 30 minutes per day through AI assistance.
Despite some reluctance at moving forwar employers seem to be taking note of AI skills, with two-thirds saying they would not hire a candidate without any experience in this domain. In fact, 71% would opt for a less experienced applicant with AI capabilities over someone more seasoned but lacking those skills.
The report claims AI usage has nearly doubled in just the last six months, it’s clear this phenomenon is rapidly going mainstream in the workplace. As one researcher put it, “Employees want AI and they aren’t going to wait. The hard part is going to be ensuring its responsible, strategic use by companies.”
We get a lot of these reports, some of which are of marginal value. This is one of the exceptions and we have a link to the full report in the show notes at technewsday.ca
Sources include: Microsoft and LinkedIn report.
Another report that caught our attention was on a related topic, this time from search firm Robert Half, another firm that we regard as developing credible research. In theire report Building Future-Forward Tech Teams, they have some key findings:
- Skills Gaps Become More Apparent. More than half of technology leaders surveyed (52 per cent) said they have a skills gap within their department. 48 per cent feel the skills gap has a greater impact today compared to a year ago.
- Skills Gaps Threaten Business Priorities. A limited pool of available talent and significant technology skills gaps could place priority projects in peril and risk negatively impacting business needs. Nearly all tech leaders (92 per cent) report challenges finding skilled talent, and nearly six in 10 (58 per cent) predict a significant hiring challenge will be the lack of applicants with the skill sets needed to support essential initiatives.
- AI Brings Skills Gaps into Focus. The need to solve skills gaps has accelerated with the emergence of new AI capabilities. As shown in the report, 79 per cent of tech leaders plan to implement initiatives involving AI this year, and 47 per cent cite a lack of staff with AI skills as the biggest barrier to success. In fact, AI and machine learning topped the list of skills where tech managers said gaps are most evident, with more than half (52 per cent) reporting this for their department.
- Building Future-Forward Tech Teams. To help hire, retain and advance high-performing tech teams, Robert Half recommends the strategies of upskilling, seeking and developing high-potential candidates, and embracing a scalable talent model.
Sources include: Building Future-Forward Tech Teams
And for a classic lesson on how not to manage employees.
Dell Technologies is rolling out a new system to monitor and grade hybrid employees based on how often they come into the office.
Starting next week, the computer giant plans to track employee badge swipes and VPN usage to determine their level of on-site presence each quarter. Employees will then be assigned a color-coded rating:
Blue = Consistent on-site presence
Green = Regular on-site
Yellow = Some on-site
Red = Limited on-site
The program is part of Dell’s push for hybrid workers to be in the office at least 39 days per quarter – essentially three days a week on average.
While managers seem to have differing interpretations, sources say the implication is that employees need to maintain a blue or green status to meet expectations. Red flags could potentially hinder career growth and make employees more vulnerable in layoffs.
Dell says the in-office requirement aims to “drive innovation and value” by combining flexible remote days with “critical” in-person collaboration.
The return-to-office mandate has been contentious since it was first announced in February. At that time, Dell stated remote workers would face disadvantages like limited career mobility.
However, multiple Dell insiders told The Register the tracking seems unusually strict for a company that previously allowed more flexibility around office culture and schedules.
Critics are slamming the policy as overly aggressive monitoring akin to “going back to grade school.” They allege the real intent is to drive workforce attrition through indirect pressure. One referred to it as “Dell on earth.”
While details are unclear, sources indicate Dell may use the data to facilitate further layoffs in addition to the at least 13,000 employees let go already this year.
The company is navigating an economic slowdown, reporting revenue down 14% to $88 billion for fiscal 2024 amid softening tech demand.
Sources include: The Register
You might remember the story on the “im-good-gpt-2-chatbot” model appeared briefly on the LMSYS benchmarking site before disappearing, only to resurface again recently. This prompted speculation that it could be an early test version of GPT-5 from OpenAI.
OpenAI CEO Sam Altman added fuel to the fire with some cryptic posts on X (formerly Twitter) referencing “gpt2” and the chatbot’s name, leaving people wondering if OpenAI was behind this mysterious model.
Some who managed to get some time to evaluate the model have posted some of the findings.
The “im-good-gpt-2-chatbot” model has sparked intense curiosity and testing from AI developers and enthusiasts ever since its appearance on the LMSYS benchmarking site. Their findings reveal some impressive – and puzzling – capabilities of this mysterious model.
One user, Min Choi, was able to create a fully functional Flappy Bird game clone using just a simple prompt and three images fed to the chatbot. This showcases its strong multi-modal skills in understanding natural language instructions alongside visual data.
Perhaps more remarkably, the model succeeded in solving a challenging freshman-level physics problem that even the powerful GPT-4 could not crack. This hints at enhanced reasoning abilities over current leading language models.
However, the chatbot’s performance is not consistently superior. While it retrieves relevant information rapidly, its text generation speed of around 250,000 tokens per minute is actually quite slow compared to modern AI benchmarks and OpenAI’s own models.
One screenshot circulating online claims to show the model accessing data directly from OpenAI’s website – raising eyebrows about its possible affiliation with the AI research company.
Overall assessments of its general capabilities are mixed. The AI researcher Sully stated the model is “marginally better than GPT-4” based on their evaluations so far. Others speculate it could be a scaled-down teaser of GPT-5 capabilities.
But this preview, if that’s what it is, reveals some impressive problem-solving and multi-modal skills that could foreshadow major advances in OpenAI’s upcoming AI models, like GPT-5. The chatbot’s training data and techniques are currently a mystery.
If this is a preview of GPT-5, it must be an early version. Altman has repeatedly said GPT-4 will soon be the “worst” of OpenAI’s models, so we should expect a major difference when OpenAI releases GPT-5. So if this new mysterious chat bot is from OpenAI, it probably isn’t the full GPT-5. But the possibility of OpenAI just doing some early stress-testing of future models or features like multi-modal and search can’t be ruled out.
And that’s our show.
Hashtag trending goes to air five days a week with a weekend interview show. We’re on YouTube,with our video version and an experimental video version where you can see me reading the news.
Visit us at our new home at technewsday.ca or .com – you pick. And you can reach me with comments, suggestions or even criticism at at editorial@technewsday.ca
You can now find our show notes at on the podcast page of technewsday.ca and even play the latest shows there, or on the home page.
